This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] Speed up for_each_rtx
- From: Jakub Jelinek <jakub at redhat dot com>
- To: Michael Matz <matz at suse dot de>
- Cc: Paolo Bonzini <paolo dot bonzini at lu dot unisi dot ch>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 20 Jan 2005 11:56:01 -0500
- Subject: Re: [PATCH] Speed up for_each_rtx
- References: <41EFAC3B.email@example.com> <Pine.LNX.firstname.lastname@example.org>
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
On Thu, Jan 20, 2005 at 05:52:00PM +0100, Michael Matz wrote:
> > (I'll point out once more that a long time in bootstrap is spent in ld
> > -- about 8 minutes).
> You need enough memory so that the .o files for libjava can be cached
> (i.e. while linking there shouldn't be too much head thrashing, otherwise
> you have not enough).
> And you need not-too-old versions of binutils, otherwise the merging of
> strings (of which there are many for debuggable objects and even without)
> is quadratic.
> For me libgcj.so links in about 24 seconds.
And you also want to avoid -Wl,-O1 when linking libgcj.so.
That is quadratic in number of dynamic symbols and libgcj.so has many
thousands of them.